The most muscular compiler switch I ever saw

[Dann Corbit]

I exited the program after each run.
C:\Users\dannc\chess\Stockfish-master2\src> stockfish-x64-native.exe
Stockfish dev-20260204-nogit by the Stockfish developers (see AUTHORS file)
Stockfish dev-20260204-nogit by the Stockfish developers (see AUTHORS file)
setoption name SyzygyPath value f:\Syzygy
info string Found 747 WDL and 741 DTZ tablebase files (up to 7-man).
bench 32767 30 20
Total time (ms) : 149619
Nodes searched : 3474345903
Nodes/second : 23221288

Total time (ms) : 196277
Nodes searched : 5294054632
Nodes/second : 26972363

Total time (ms) : 161237
Nodes searched : 5234443773
Nodes/second : 32464284

Total time (ms) : 135235
Nodes searched : 3778092928
Nodes/second : 27937242

C:\Users\dannc\chess\Stockfish-master2\src> stockfish-x64-avx2.exe
Stockfish dev-20260204-nogit by the Stockfish developers (see AUTHORS file)
Stockfish dev-20260204-nogit by the Stockfish developers (see AUTHORS file)
setoption name SyzygyPath value f:\Syzygy
info string Found 747 WDL and 741 DTZ tablebase files (up to 7-man).
bench 32767 30 20
Total time (ms) : 219029
Nodes searched : 6022852879
Nodes/second : 27497970

Total time (ms) : 214674
Nodes searched : 6522724886
Nodes/second : 30384326

Total time (ms) : 289362
Nodes searched : 9207555532
Nodes/second : 31820195

Total time (ms) : 154100
Nodes searched : 4508083382
Nodes/second : 29254272

I do see that there is a lot of randomness here, and that the last run for avx2 was faster than two of the runs for native.
I guess I was comparing the 135 second run to the 289 run, in my surprising case.
And we should not be surprised that the native version wins more often since it is a little faster single threaded.